Premium
QSTR with Extended Topochemical Atom Indices. 10. Modeling of Toxicity of Organic Chemicals to Humans Using Different Chemometric Tools
Author(s) -
Roy Kunal,
Ghosh Gopinath
Publication year - 2008
Publication title -
chemical biology and drug design
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.59
H-Index - 77
eISSN - 1747-0285
pISSN - 1747-0277
DOI - 10.1111/j.1747-0285.2008.00712.x
Subject(s) - partial least squares regression , quantitative structure–activity relationship , atom (system on chip) , chemistry , biological system , set (abstract data type) , computational chemistry , computer science , stereochemistry , machine learning , programming language , biology , embedded system
In this communication, we have developed quantitative predictive models using human lethal concentration values of 26 organic compounds including some pharmaceuticals with extended topochemical atom (ETA) indices applying different chemometric tools and compared the extended topochemical atom models with the models developed from non‐extended topochemical atom ones. Extended topochemical atom descriptors were also tried in combination with non‐extended topochemical atom descriptors to develop better predictive models. The use of extended topochemical atom descriptors along with non‐extended topochemical atom ones improved equation statistics and cross‐validation quality. The best model with sound statistical quality was developed from partial least squares regression using extended topochemical atom descriptors in combination non‐extended topochemical atom ones. Finally, to check true predictability of the ETA parameters, the data set was divided into training ( n = 19) and test ( n = 7) sets. Partial least squares and genetic partial least squares models were developed from the training set using extended topochemical atom indices and the models were validated using the test set. The extended topochemical atom models developed from different statistical tools suggest that the toxicity increases with bulk, chloro functionality, presence of electronegative atoms within a chain or ring and unsaturation, and decreases with hydroxy functionality and branching. The results suggest that the extended topochemical atom descriptors are sufficiently rich in chemical information to encode the structural features for QSAR/QSPR/QSTR modeling.